Volcanoes deform as a consequence of the rise and storage of magma; once magma reaches a critical pressure, an eruption occurs. However, how the edifice deformation relates to its eruptive behavior is poorly known. Here, we produce a joint interpretation of spaceborne InSAR deformation measurements and volcanic activity at Mt. Etna (Italy), between 1992 and 2006. We distinguish two volcano-tectonic behaviors. Between 1993 and 2000, Etna inflated with a starting deformation rate of ∼1 cm yr−1 that progressively reduced with time, nearly vanishing between 1998 and 2000; moreover, low-eruptive rate summit eruptions occurred, punctuated by lava fountains. Between 2001 and 2005, Etna deflated, feeding higher-eruptive rate flank eruptions, along with large displacements of the entire East-flank. These two behaviors, we suggest, result from the higher rate of magma stored between 1993 and June 2001, which triggered the emplacement of the dike responsible for the 2001 and 2002–2003 eruptions. Our results clearly show that the joint interpretation of volcano deformation and stored magma rates may be crucial in identifying impending volcanic eruptions.
 Understanding the eruptive behavior of a volcano is crucial to minimize its hazard. The eruptive behavior depends upon when (frequency), where (which portion) and how (eruptive style and rate) a volcano erupts its magma. As specific behaviors are accompanied by deformation of the volcanic edifice and portions thereof, the volcanic and structural evolutions are closely coupled. Modern geodetic techniques, that allow the prompt detection of millimeter-scale deformation, may be invaluable for forecasting volcanic eruptions. However, applications are currently limited to particular episodes [Amelung et al., 2007] and do not permit defining the general behavior of the entire volcanic edifice over long time scales.
 To understand how volcanoes deform during the rise, storage and eruption of magma, we consider 15 years of spaceborne interferometric synthetic aperture radar (InSAR) data covering the eruptive activity of Mt. Etna, between 1992 and 2006 (Figure 1). This basaltic volcano is one of the largest and most active worldwide, erupting every ∼1.5 years over the last century [Behncke et al., 2005]. Eruptions may occur from the summit craters (summit eruptions) and/or from the slopes (flank eruptions) (Figure 1b). The latter are mainly fed by dikes propagating laterally from the central conduit and, exceptionally, by dikes propagating vertically, independently of the central conduit [Acocella and Neri, 2003]. A typical cycle of activity, with a time span of decades, as visible at the surface, begins with degassing from the summit craters. This is followed by progressively increasing summit eruptions activity and then by voluminous flank eruptions, with sporadic summit activity [Behncke and Neri, 2003].
 Based on the deformation measurements retrieved from our InSAR analysis and on the volcanological data recorded during the 1992–2006 time interval, we investigate the volcanic and structural behavior of Mt. Etna during the last eruptive cycle.
 The key point of our study is represented by the joint interpretation of the volcanic activity of Etna and of its deformation retrieved by using InSAR data.
 The volcanic activity analysis is based on investigating the cumulative erupted and degassed magma volumes between 1993 and 2006; in particular, the latter are inferred from bulk SO2 emissions from the whole active craters, considering a magma density of 2.65 kg m−3 and a mean S content equals to 0.27% [Allard et al., 2006]. The measurements were acquired two times per week, interpolating the data, with a mean error of 25% [Bruno et al., 2001].
 The InSAR analysis relies on the Small BAseline Subset (SBAS) approach [Berardino et al., 2002; Pepe et al., 2005] to resolve the space-time evolution of the surface deformation. In particular, we process 107 ascending (Track 129, Frame 747) and 102 descending (Track 222, Frame 2853) SAR data acquired by the ERS-1/2 and ENVISAT sensors, for the 1992–2006 period. From these SAR data we compute 283 interferograms from the ascending orbits and 289 from the descending ones; these are subsequently inverted by applying the SBAS-InSAR technique. At this stage, the availability of InSAR deformation time series relevant to both the ascending and descending radar line-of-sights (LOS) enables separation of the mostly Vertical (V) and East–West (E–W) components of the displacements [Manzo et al., 2006]. This procedure is carried out on a year to year basis (see auxiliary material). This dataset constitutes an exceptional opportunity to monitor the long-term behavior of a volcano, as well as a significant improvement with regard to the previous Mt. Etna InSAR studies, focused on deformation maps retrieved from single interferograms [Massonnet et al., 1995; Borgia et al., 2000; Froger et al., 2001; Lundgren et al., 2003; Lundgren and Rosen, 2003; Neri et al., 2007] or investigating InSAR time series limited to the 1992–2000 interval [Lundgren et al., 2004].
3. Recent Volcanic Activity
 The onset of the last eruptive cycle started after the end of the 1991–1993 flank eruption [Allard et al., 2006]. The volcano degassed (April 1993–June 1995), then showed almost continuous summit activity, with increasing intensity from July 1995 through June 2001, culminating in flank and summit eruptions from July 2001 through 2006 (Figure 1b). Between 1995 and 2001, ∼1.2 × 108 m3 of magma were erupted [Allard et al., 2006] by means of longer (months) effusive eruptions, with low to moderate rates and shorter (hours) lava fountains, with higher eruptive rates. During this active period (1982 days), the mean rate of erupted magma was ∼0.7 m3 s−1. Between July 2001 and 2006, ∼1.5 × 108 m3 of magma were erupted by a few events with higher rates [Allard et al., 2006; Behncke et al., 2008]; during the period of activity (465 days), these correspond to ∼3 m3 s−1 (Figure 1c). The 2001 and 2002–2003 flank eruptions, both mainly fed by a vertically propagating dike on the south flank and with the highest eruptive rates (Figure 1c), were accompanied by seismic activity and extensive surface fracturing on the east flank [Neri et al., 2005]. The cumulative magma volumes degassed and erupted between 1993 and 2006 are shown in Figure 1d.
4. InSAR Results
Figure 2 shows the time frames summarizing the vertical and horizontal (E–W component) deformation between 1992 and 2006, while the animation is provided in the auxiliary material. A deflation of the volcano followed the 1991–1993 eruption (Figure 2a) [Massonnet et al., 1995]. From 1993 to 2000, as evident from the vertical (Figures 2b–2d) and eastward and westward (Figures 2j–2l) displacement of the summit, the volcano inflated with a starting rate of ∼1 cm yr−1 that progressively reduced with time, nearly vanishing during the 1998–2000 time interval (Figures 3c–3e, see vertical displacements). Contemporaneously, the lower portions of the east and south flanks also showed significant deformation; in particular, the lower east flank shifted eastward (Figures 2j–2l), confirming the previously suggested seaward slip [Borgia et al., 1992; Froger et al., 2001]; however, also in this case the amount of East flank motion showed a significant slowdown during the 1998–2000 interval (Figure 2l).
 The 2001 flank eruption brought about major changes to the volcano deformation behaviour. During this eruption (mean eruptive rate ∼15 m3 s−1; Figure 1c), the horizontal deformation on the west and east flanks increased by at least one order of magnitude (Figures 2m and 3c–3d, see E–W displacements): the upper portions of the west and east flanks underwent a westward and eastward displacement of several tens of cm respectively. This resulted from the emplacement of the N–S trending dike feeding the eruption on the upper South Rift [Bonaccorso et al., 2002] and spreading of the edifice [Lundgren and Rosen, 2003].
 The large displacements culminated in the 2002–2003 eruption (Figures 2f and 2n). With a mean eruptive rate ∼6 m3 s−1 (Figure 1c), it was fed by the same N–S trending dike of the 2001 eruption and, subordinately, by a dike along the NE Rift [Neri et al., 2005]. The emplacement of these dikes increased the horizontal and vertical deformation of the upper flanks. The horizontal deformation increased by several tens of cm and shifted, with respect to the 2001 displacements (Figure 3d), northward (Figure 3e), northeastward (Figure 3f) and eastward (Figure 3g) with large displacement along the Pernicana Fault System (PFS; Figure 1), decreasing southward.
 Between 2001 and 2003, deflation phenomena (6–13 cm) affected most of the volcano summit (Figures 2e–2f and 3c–3e). The comparison between the InSAR and previous structural field data (Figures 3a–3b) [Allard et al., 2006] shows that, in this period, while the deformation on the west flank was not associated with any fault, the deformation on the east flank was largely fault-controlled. These faults induced seismic activity and surface fracturing [Acocella et al., 2003; Neri et al., 2004].
 After 2003, the 2004–2005 and the 2006 eruptions, both associated with laterally propagating dikes, were characterized by eruptive rates of ∼2.5 m3 s−1 [Neri and Acocella, 2006; Behncke et al., 2008], notably lower than those of the two previous flank eruptions. Most of the deformation consisted of the eastward shift of the east flank of the volcano, coupled with the deflation of the volcano summit (Figures 2g–2h and 3), interrupted by evidence of new uplift during the 2005–2006 time interval.
 Finally, the anticline located in the lower south flank, likely induced by compression at the edifice base [Borgia et al., 1992, 2000], showed a long-term uplift rate of ∼0.7 cm yr−1 (Figures 2 and 3h) with some triggering effects due to the recharging of the magma system [Lundgren et al., 2004].
5. Discussion and Conclusions
 The 1992–2006 InSAR and volcanological data allow us to understand the behavior of Etna during its last cycle of activity. To effectively represent these data and simplify their interpretation we present in Figure 4 the temporal evolution of the mean vertical displacement of the volcano summit (computed within the area highlighted by the white box in Figure 3a) and of the magma stored below the volcanic edifice. The former synthesizes well the average deformation characteristics (see Figures 3c–3f); the latter is computed as the difference between the degassed and the erupted magmas (Figure 1d). The joint interpretation of the InSAR and volcanological data allows us to distinguish two volcano-tectonic behaviors associated to the 1993–2000 and 2001–2005 time intervals, respectively.
 With respect to the former interval, a general inflation affected the volcano from 1993 to 1998, without volumetrically significant eruptions from the summit, while from 1998 to 2000 this inflation drastically decreased, probably due to the increase of the eruptive activity (Figure 1d). This deformation behavior was accompanied, until June 2001, by high rates of stored magma (∼1.2 108 m3 yr−1) leading to an overall storage of ∼9.7 × 108 m3 (Figure 4). This significant amount of magma emplaced in a relatively short period may have triggered the vertically propagating and eccentric dike responsible for the 2001 and 2002–2003 eruptions [Neri et al., 2005]. This dike, providing an additional path for the rise and emission of magma, is an exceptional event in the recent history of Etna [Acocella and Neri, 2003]. Its emplacement caused important variations in the eruptive and deformative volcano styles (Figure 4).
 With respect to the second time interval, the volcanic activity was more concentrated in time between 2001 and 2003, with significantly higher eruptive rates (6–15 m3 s−1; Figure 1c) and deflation phenomena that affected the volcano summit. Moreover, from late 2003, when the eccentric dike became inactive, volcanic activity and displacement rates decreased but the deflation of the volcanic edifice was still significant. The lower rates of stored magma (3.7 107 m3 yr−1; see Figure 4), recognized from July 2001 to the end of 2006 appear ultimately related to the decrease in the magmatic supply (degassed magma curve; Figure 1d); these rates match with the observed deflation phenomena affecting the volcanic edifice, due to the volcanic and structural paroxysms related to the significant drainage of the stored magma.
 These results clearly show that the joint interpretation of volcano deformation and stored magma rates may be crucial for identifying impending volcanic and tectonic eruptions. Indeed, our analysis highlights that the presence of ongoing inflation with no reduction of the stored magma rate yields conditions that lead to the occurrence of large eruptions.
 The 1993–2006 eruptive activity of Mt. Etna is a key example of how a constant rate of magma accumulation may lead to eruptive and structural paroxysms in an active basaltic volcano. Our study shows in detail how the deformation of an edifice relates to its eruptive behavior over a time span much larger than that of a single eruption and, as such, it offers a reference model to better understand the behavior of volcanoes in general. A rigorous assessment of this model would require numerical models relating the balance between pressure accumulation, rock strength and dike propagation, which is beyond the scope of this manuscript but is the next step for future analysis.
 ESA provided the SAR data (Cat-1, N. 3560). The DEM was obtained from the SRTM archive while the ERS-1/2 orbits are courtesy of the TU-Delft, Netherlands. This work was partly funded by INGV and the Italian DPC and was supported by ASI, the Preview Project and CRdC-AMRA. F. Amelung, M. E. Pritchard and an anonymous reviewer provided very useful suggestions. We acknowledge the DPC-INGV Flank project for providing the funds for the publication fees. Moreover, a portion of the research described in this paper was supported under contract with the National Aeronautics and Space Administration at the Jet Propulsion Laboratory.